During the course of expressing the disorazol gene cluster, B. thailandensis E264 showed better performance than the well-studied host DSM 7029 (Bian et al., 2017; Yan et al., 2018; Liu J. et al., 2021) in terms of growth rate and production. To facilitate genetic manipulation, three efflux pumps in strain E264, AmrAB-OprA (abbreviated as A), BpeAB-OprB (B), and BpeEF-OprC (C) (Biot et al., 2011, 2013), were inactivated by homologous recombination combined with pheS counterselection mediated seamless gene deletions (Barrett et al., 2008), and mutants with one, two, or three pump knockouts were obtained. Antibiotic susceptibility determined by the disc diffusion test was shown in tables S4 and S5. For most antibiotics, the mutant with one pump knockout (E264ΔA, E264ΔB, or E264ΔC) showed improved antibiotic sensitivity compared to the wild-type, suggesting that they are involved in multidrug resistance. Therefore, the mutant with three pump knockouts (E264ΔBAC) was constructed and used to express gene clusters or characterize promoters in subsequent experiments.

Compared with the wild-type, E264ΔBAC was more sensitive to the tested antibiotics, except ampicillin (Figure 2A, Table 1). The minimum inhibitory concentrations (MIC) of the tested antibiotics against mutant E264ΔBAC were much lower than the concentrations used to select transformants (Supplementary Table S6). The MIC value (<7.5 μg/ml) of kanamycin against the mutant was much lower than the concentration (300 μg/ml) used to select transformants of the wild type (Kang et al., 2011). The MIC values of tetracycline, chloramphenicol, and erythromycin are lower than 0.6, 1.9, and 3.1 μg/ml, respectively. The antibiotic-sensitive mutant E264ΔBAC facilitates the genetic engineering of this potential heterologous host with low-concentration antibiotics as a selectable marker. In addition, the growth rate of the mutant E264ΔBAC was not affected by engineering, and the growth curve of the mutant was similar to that of the wild-type strain E264, except that the maximum OD600 was slightly lower (Figure 2B).

After obtaining the mutant, the plasmid p15A-dis (Supplementary Figure S1; Tu et al., 2016) was introduced into the mutant E264ΔBAC via conjugation mediated by the donor strain Escherichia coli WM3064 (Dehio and Meyer, 1997). A specific peak at m/z 729.4 was observed in the crude extract of the mutant with the disorazol gene cluster, and the area of the specific peak produced by mutant E264ΔBAC was much higher than that of S. brevitalea DSM 7029, which we used for heterologous expression of cis-PKS-derived epothilone (Figure 1B and Supplementary Figure S2; Bian et al., 2017), suggesting that E264 is more suitable for expressing this trans-AT PKS gene cluster. However, the molecular weight of the product (1) was different from that of disorazol A₁ (Figure 1B). The molecular formula of 1 was predicted to be C₄₂H₅₂N₂O₉ based on HR-ESIMS (m/z 729.3773 [M + H]⁺, calcd 729.3746; Figure 1B), which is consistent with disorazol F₂. Furthermore, the ¹H and ¹³C NMR spectra of 1 confirmed that the product is disorazol F₂ (1) (Supplementary Table S7, Supplementary Figures S3 and S4; Jansen, 1994). Compared to disorazol A₁, the epoxide at C-9/C-10 in 1 was replaced by a double bond, and the hydroxyl group at C-6′ was not methylated (Figure 1A).

The heterologous expression of disorazol in a phylogenetically distant relative of the myxobacterial strain S. cellulosum could exclude the influence of homologous genes in the previously reported myxobacterial host M. xanthus DK1622 (Tu et al., 2016) and facilitate the elucidation of its biosynthetic pathway. The orf9 encoding a ubiquinone biosynthesis protein Coq4 is not necessary for disorazol production (Tu et al., 2016). The product produced by the heterologous host B. thailandensis E264 is disorazol F₂ rather than A₁ or A₂ produced by original producer So ce12 or myxobacterial host DK1622, indicating that the epoxidation and methylation of disorazols may be catalyzed by enzymes encoded by genes located outside of the gene cluster. Another debate regarding disorazols is the installation of a hydroxyl group at C-6. Theoretically, the DH7 domain could catalyze the elimination of hydroxyl groups at C-6 and C-6′ and form symmetrical macrodiolide products; however, most natural disorazols are unsymmetrical owing to the presence of the hydroxyl group at C-6′. The product obtained from E264ΔBAC::Ptet-dis is disorazol F₂, an unsymmetrical macrodiolide, which suggests that the hydroxyl group at C-6′ is not installed during post-PKS modification.

Initially, the yield of disorazol F₂ (1) was less than 0.4 mg/L. To improve the yield of F₂ (1), we first characterized potent native promoters in E264. Fifty native promoters with high expression levels at different phases of growth in M9 broth were selected based on the data of transcriptome analysis. However, the data of transcriptome may be affected by sample processing procedures, such as liquid nitrogen freezing, we further characterized the promoters in vivo by a luciferase assay to obtain a series of potent constitutive promoters of E264 (Supplementary Figure S5, Supplementary Table S8; Ouyang et al., 2020). To avoid possible positional effects, an attB site, the attachment site of site-specific recombination, was integrated into the locus of operon BpeEF-OprC, resulting in the mutant E264ΔBAC::attB. The potent promoters P11 (promoter of a hypothetical protein), P17 (promoter of a radical SAM protein), P33 (promoter of a hypothetical protein), P35 (promoter of a peroxiredoxin), P44 (promoter of a membrane protein), and P46 (promoter of a DUF4148 domain-containing protein) were employed to promote the expression of the disorazol gene cluster, resulting in a significant increase in yield (Figure 3A). Among them, the optimal promoter, P46, resulted in an approximately 23-fold increase in the average yield of disorazol F₂ (1) to 9.3 mg/L. Promoters P17 and P44 showed similar effects, which led to an obvious increase in yield to 7.5 and 8.6 mg/L, respectively. Other promoters also improved the yields by 1.5-fold to 6.5-fold. To further improve the yield, promoters P46, P44, P17, and P46 were inserted upstream of genes disA, disB, disC, and disD using the RedEx method (Supplementary Figure S6; Song et al., 2020), which generated mutants 1P (P46), 2P (P46 + P44), 3P (P46 + P44 + P17), and 4P (P46 + P44 + P17 + P46; Figure 3B). The yield of disorazol F₂ (1) showed a positive correlation with the number of promoters, which reached to 38.3 mg/L in mutant 4P-dis, 96-fold higher than that of E264ΔBAC::Ptet-dis. The significantly improved efficiency of heterologous expression encouraged us to express BGCs from other phylogenetical taxa.

In addition to the gene cluster from myxabacteria, the BGCs of rhizoxin (rhi) from Burkholderia rhizoxinica (Partida-Martinez and Hertweck, 2007) and shuangdaolide (sdl; Supplementary Figure S7) from Streptomyces sp. B59 (Liu Y. et al., 2021) were transferred to E264ΔBAC::attB. No specific peak was detected in the crude extract of E264ΔBAC::attB/sdl, whereas a series of specific peaks were observed in the fermentation product of E264ΔBAC::attB/rhi, whose molecular weights were consistent with the reported rhizoxins (Figure 4). Specific peaks a and b with a high proportion were detected at m/z 628.3 with t_R = 13.4 min and at m/z 642.3 with t_R = 14.8 min. Based on HRMS, the molecular formula of the compound in peak a was established as C₃₅H₄₉NO₉ (m/z 628.3483 [M + H]⁺, calcd 628.3482), which is identical to that of rhizoxins M₁, Z₂, and S₂; peak b may contain rhizoxin M₂ or Z₂ (C₃₆H₅₁NO₉, m/z 642.3636 [M + H]⁺, calcd 642.3637). In addition to the major peaks, we also observed some specific peaks in a relatively low proportion of the fermentation product of E264ΔBAC::attB/rhi. Their MS/MS spectra were similar to peaks a and b, and HRMS suggested that they were also related to known rhizoxins (Supplementary Table S9). The BGC of rhizoxin from closely related species was expressed successfully in the host based on strain E264, and the potential of this heterologous expression system still needs to be further developed.

Owing to the low yield of the original producer So ce12, the cytotoxicity of disorazol F₂ has never been reported in previous studies. In this study, the cytotoxicity of disorazol F₂ (1) was evaluated against three human cancer cell lines (HepG2, HCT116, and A549; Table 2). Although one-two orders of magnitude weaker than epoxide disorazols A₁ and A₂ (Tu et al., 2016), disorazol F₂ still exhibited potent activity against the tested cell lines with IC₅₀ values ranging from 0.25 to 3.11 nM, one to three orders of magnitude stronger than clinically used doxorubicin or vincristine. The cytotoxicity assay compared with previously reported data (Irschik et al., 1995; Elnakady et al., 2004; Hopkins and Wipf, 2009; Tu et al., 2016) suggested that the activity of disorazols is affected by the epoxide at C-9/C-10, which suggests the tailoring enzyme of the disorazol pathway is still a significant task.

Burkholderia thailandensis E264 was purchased from German collection of microorganisms and cell cultures GmbH (DSMZ) and was cultured in low-salt Luria-Bertani medium (1% tryptone, 0.5% yeast extract, 0.1% NaCl) at 37°C. The markerless mutants of strain E264 were established as described in section construction of deletion mutants. Escherichia coli GB05-red was used for mediating homologous recombination between a linear and circular DNA molecular (LCHR) while E. coli GB05-dir mediating linear plus linear recombination (LLHR). Escherichia coli GB05-dir-gyrA462 or GB05-red-gyrA462 were used for constructing plasmids with counterselectable marker CcdB. Escherichia coli WM3064 was employed as a conjugal donor (Dehio and Meyer, 1997). Unless otherwise specified, the strains of Escherichia coli were cultured under the same conditions with strain E264. Tetracycline (Tet) was added to a final concentration of 50 μg/ml for selective growth of wild type strain of B. thailandensis E264. For mutant E264ΔBAC, antibiotics were added to final concentrations of 30 μg/ml kanamycin (Kan) or 5 μg/ml gentamicin (Genta) as required, for strains of E. coli, the concentrations were 30 μg/ml Kan, 100 μg/ml ampicillin (Amp) or 15 μg/ml chloramphenicol (Cm).

The genomic DNA (gDNA) of strain E264 was isolated using phenol – chloroform – isoamyl alcohol mixture (Wang et al., 2020). To obtain markerless mutants, genetic tools developed by Barrett et al were modified and applied in this study (Barrett et al., 2008). Briefly, about 1 kb regions upstream and downstream of targeted genes amplified by PCR using PrimeSTAR® HS DNA polymerase with GC buffer were cloned into vector pBR322-amp-tet-pheS (linearized with restriction enzymes XabI and NcoI) by strain GB05-dir. The plasmids with homology arms were introduced into strain E264 by natural transformation (Garcia, 2017), and correct clones verified by colony PCR were cultured in LSLB without antibiotic for 12 h to eliminate selection marker. The overnight cultures were centrifuged and washed twice with liquid M9G medium (M9 medium of Sangon, Shanghai supplemented with 20 mM Glucose), then 50 μl cell suspensions were plated on M9G plates supplementing with 0.1% 4-Chloro-D, L-phenylalanine (Sigma-Aldrich) to screen the recombinants with selection marker deletion. Markerless mutants were further verified by colony PCR. To generate mutant E264ΔBAC::attB, the sequence of attB (GGGTGCCAGGGCGTGCCCTTGGGCTCCCCGGGCGCGTA) was inserted between the homology arms of pBR322-Amp-Tet-pheS-oprC through oligonucleotide synthesis, and the plasmid was introduced into mutant E264ΔBA followed by selection and counterselection successively.

To make test plates, the overnight cultures (200 μl) of strain E264 or E264ΔBAC were mixed with 50 ml melted LSLB agar medium. Then, the paper discs containing different antibiotics (HangweiTM, Hangzhou) were placed on the media. The plates were cultured at 37°C overnight to measure the diameter of inhibition zone.

The amp-ccdB cassette flanked with two restriction enzyme NdeI sites and about 40 bp homology arms was amplified from plasmid p15A-amp-ccdB using primers amp-ccdB-P11-F/R (Wang et al., 2018), the PCR product and plasmid p15A-genta-int-attP-P11-firefly (Ouyang et al., 2020) were electroporated into E. coli GB05-red-gyrA462 expressed recombinases Redα/Redβ, which resulted in plasmid p15A-genta-int-attP-amp-ccdB-firefly. Colonies growth on plates supplemented with Amp and Genta were transferred into 1.8 ml fresh LB broth with appropriate antibiotics and incubated at 37°C overnight. The plasmid DNA was extracted from the overnight culture and confirmed by restriction analysis.

Promoter Px was amplified using primers PX-F/R with 40 bp homology arms, and plasmid p15A-genta-int-attP-amp-ccdB-firefly was linearized with restriction endocuclease NdeI. The products of PCR and restriction enzyme digestion were electroporated into E. coli GB05-dir expressed recombinases RecE/RecT, which resulted in plasmid p15A-genta-int-attP-Px-firefly.

Plasmids with different promoters located upstream of luciferase gene were introduced into E264ΔBAC::attB by conjugation mediated by WM3064 (Dehio and Meyer, 1997), The expression levels of report gene in transformants were detected by Luciferase Assay System of Promega Corporation. Briefly, 20 μl cell lysates prepared according the instruction were mixed with 100 μl of Luciferase Assay Reagent, and the light produced by the mixture was quantified by GloMaxTM 96 Microplate Luminometer of Promega Corporation.

Cloning vector BAC-cm flanked with PacI/HpaI sites and homology arms was amplified by PCR from plasmid pBeloBAC11 (digested with BamHI and HindIII) using primers bac-dis-F/R. The products of PCR and plasmid p15A-dis (Tu et al., 2016) linearized with XbaI were electroporated into E. coli GB05-dir, which resulted in plasmid pBAC-cm-dis.

Plasmid p15A-genta-int-attP-Px-firefly was linearized with MfeI, then cassette p15A-genta-int-attP-Px with homology arms was amplified by PCR using primers Px-dis-F/R. The PCR product and linearized plasmid pBAC-cm-dis (digested with PacI and HpaI) were assembled by LLHR mediated by GB05-dir, which resulted in plasmid p15A-genta-int-attP-Px-dis.

To construct plasmids with multiple promoters and improve the yield of disorazol, the strategy of RedEx was employed (Song et al., 2020). Breifly, plasmid p15A-genta-int-attP-P46-dis was electroporated into E. coli GB05-red-gyrA462, then the cassette 2P-amp-ccdB, dis (17985–18044)-PacI-amp-ccdB-PacI-dis (18015–18044)-P44-dis (18053–18110; the first nucleotide of disA was designated as 1), generated by fusion PCR was electroporated into E. coli GB-red-gyrA462 with plasmid p15A-genta-int-attP-P46-dis. The cassette 2P-amp-ccdB was inserted upstream of disB through LCHR mediated by recombinases Redα/Redβ which resulted in plasmid p15A-genta-int-attP-2P-dis-amp-ccdB. The plasmid p15A-genta-int-attP-2P-dis-amp-ccdB was digested with PmeI, and 100 ng linearized plasmid DNA was treated with 0.2 U T4pol (New England BioLabs) in a 20 μl reaction at 25°C for 30 min, 75°C for 20 min, 50°C for 20 min and then held at 4°C in a thermocycler. The reaction mixture was electroporated into E. coli GB2005 cells after desalting treatment. The recombinant p15A-genta-int-attP-2P-dis (2 promoters) with promoter P44 inserting upstream disB was identified with KpnI/PmeI restriction analysis.

Utilizing the strategy of RedEX, promoter P17 was inserted upstream of gene disC and P46 was inserted upstream of gene disD successively, which resulted in plasmid p15A-genta-int-attP-3P-dis (3 promoters) and p15A-genta-int-attP-4P-dis (4 promoters), respectively.

Single colonies were inoculated into 2 ml Eppendorf tubes containing 1.8 ml LB broth supplemented with appropriate antibiotics and incubated at 37°C overnight with shaking at 900 rpm. The seed cultures (1%, v/v) were transferred into 50 ml M9 broth in 250 ml Erlenmeyer flask and incubated at 30°C with shaking at 200 rpm for 24 h before 2% XAD-16 resin was added, then the incubation was continued for another 48 h. The resin was collected by centrifugation and resuspended with 50 ml of methanol. The mixtures were shaken at 30°C, 200 rpm for 2 h. The methanol was removed by evaporation in vacuo, and residues were dissolved with 1 ml of methanol. After filtering with 0.22 μm membrane, the crude extracts were analyzed by UPLC-MS (UltiMate 3,000 UPLC system combined with Bruker amazon SL Ion Trap mass spectrometer). The C18 column (2.1 × 100 mm. 2.2 μm, Thermo) was utilized to analyze the crude extracts at a 0.3 ml/min flow rate using the following program: 0–3 min 5% solvent B; 3–22 min, 5 95% with linear gradient; then, 22–25 min, 5% solvent B (Solvent A, Milli Q water supplemented with 0.1% formic acid; Solvent B, acetonitrile supplemented with 0.1% formic acid). Mass spectra were acquired in positive ion mode.

The crude extract from 10 L culture of E264ΔBAC::Ptet/dis was fractionated with Sephadex LH-20 column (GE Healthcare) chromatography using MeOH as a mobile phase. Fractions containing F₂ (1) were combined and further purified by semipreparative reverse-phase HPLC (Agilent ZORBAX SB-C18 column, 250 × 9.4 mm, 5 μm; gradient elution 0–2 min, 64% ACN; 2–17 min, 64–90% ACN; 17–22 min, 95% ACN, 22–25 min, 64% ACN; 2.5 ml/min). Finally, 1.5 mg of disorazol F₂ (1) was obtained with retention time at 15.1 min.

The standard curve of disorazol F₂ (1) was established by measuring the area of absorption spectrum at UV 280 nm containing 0.05, 0.1, 0.2, 0.5 and 1.0 mg/ml disorazol F₂ (1). The yield of disorazol F₂ (1) was determined by reference to the standard curve.